Al Gore Rhythms [R for BE/BA]

posted by Helmut Homepage – Vienna, Austria, 2009-09-29 14:52  – Posting: # 4269
Views: 30,932

Dear D Labes!

» Hope you have already waited for me.

Yesss!

» Here are the results of the SAS Proc Power.
» Seems the power is calculated in this case totally wrong!

Fascinating.

» My own rolled code (relying on Owens Q-functions, undocumented but available in SAS) gives:

That’s the more interesting part (homebrew)!

                        n  % power    n  % power
------------------------------------------------
#1 Diletti et al.       4 >70         6 >90
#2 PASS                 4  72.90
#3 D Labes' SAS code    4  72.9014    6  98.6992
#4 nQuery Advisor 7     4  71.559     6  97.943
#5 R-code               4  66.674     6  98.697
#6 FARTSSIE 1.6         4  66.674     6  98.697
#7 StudySize 2.0.1      4  15.9 (!)   6  93.201


n=4 takes the lead by 4:3! Observations:
#5 (based on Algorithm AS 243) and #6 give identical results. This is not surprising, because the VBA-routine within FARTSSIE was written by Russell V. Lenth, the author of AS 243. #2/3 agree; for n=6 power is comparable with #5/6. #4 calculates lower power than #1-3, but still would suggest n=4. The difference between #1-3 and #5/6 might be due to different algorithms for estimating the noncentral t-distribution (#1/3: ?, #2: AS 243?, #4: AS 184, #5/6: AS 243). #7 uses approximations (Ref. 2 in this post); however, yields conservative results.

Maybe I should contact Dieter Hauschke to join the party…

Cheers,
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

Activity
 Admin contact
20,644 posts in 4,328 threads, 1,436 registered users;
online 18 (1 registered, 17 guests [including 11 identified bots]).
Forum time: 16:44 CEST (Europe/Vienna)

No one should approach the temple of science
with the soul of a money changer.    Thomas Browne

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5